Tool Radar: A Model Picker, Smarter Spacing, Chats Side by Side

Welcome back to Tool Radar — our running note on the adjacent tools we are watching for freelancers. By "adjacent" we mean the tools that sit next to the work itself: the AI assistant you draft with, the design app you mock up in, the editor you ship from, the notes app that holds your second brain. We do not cover other freelance-management apps here — this is about the rest of your stack.

Two filters run through everything below. The first is the question this series keeps returning to: does it remove a decision, or does it just add a tab? The second is the one solo freelancers forget until it matters: what happens to your work if this tool disappears — can you get your stuff out? This week's updates each land somewhere honest on both.

AI: your notes app now helps you pick the model

Notion added a model picker on August 14 that leads with a shortlist of models made for your hardest tasks, gives each one a scorecard to compare speed, intelligence, and cost, lets you pin favorites, and adds an effort control to dial up how hard the model thinks on a single run. The release was titled, plainly, "Model selection, simplified."

Run it through the first filter and the answer is mostly encouraging. "Which model should I use for this?" is a recurring, low-confidence decision most of us have been making by vibes. A scorecard that puts the trade-off in front of you — faster versus more thorough, cheaper versus sharper — turns that into a choice you can actually reason about instead of guess. That removes a decision, or at least demotes it from research project to glance.

The one caution is the effort dial itself. A new knob is only a gift if you set it once and move on; it becomes a tab if you find yourself re-tuning it every time you draft.

• A better-chosen model still writes a draft, never a send. Whatever the scorecard steers you to, anything a client sees — a proposal, a scope note, a payment reminder — gets read line by line by you. The picker changes which engine drafts; it does not change whose name is on the result.
• Pick for the task, then leave it. Decide what "thorough" and "fast" mean for your own work once, pin the two models you actually reach for, and stop re-litigating the choice on every prompt.

This is also how we think about AI inside Drift Catch. Our two AI features — AI-Assisted Drafting on templates, and support-email triage — both produce a draft that a human reviews before anything is sent. We let the model get a first version onto the page; we never let it have the last word.

Design: edge spacing became a setting, not a nudge

Figma gave auto layout's auto spacing two new options, "Around" and "Evenly", on August 21, alongside the existing behavior now named "Between". Between puts even gaps between items and nothing at the edges; Around adds half a gap at each edge; Evenly makes every gap identical, edges included. If you think in CSS, they map to space-between, space-around, and space-evenly.

This is close to a pure example of the first filter. Getting the spacing at the edges of a row to feel right used to be a manual nudge you re-made on every layout — padding tweaked by eye until it looked even. Turning the three common patterns into named settings removes that decision: you name the behavior you want once, and the layout keeps it as things resize.

The best small features retire a chore you did not notice you were repeating.

The freelancer angle underneath is consistency. When even spacing is a setting rather than a hand-tuned value, your work reads the same across every screen and revision — one less thing a client can point at and call "off" when a project starts to drift.

Dev: compare AI chats side by side

VS Code 1.134, out August 19, lets you arrange related chats and subagent chats into groups so you can compare them side by side, adds a prompt timeline that marks each prompt in the transcript so you can jump back and see which ones changed files, and adds Find in chat to search a whole conversation. It also makes the integrated browser the default for opening local HTML files, so a preview stays inside the editor.

Through the first filter, the useful one here is the prompt timeline. When an AI-assisted session runs long, "which instruction actually changed this file?" is the question that eats your afternoon. A timeline that pins each prompt to its file changes turns that from archaeology into a click — it removes the decision about where to start looking.

The side-by-side comparison is quieter but real: seeing two attempts next to each other is how you tell which draft to keep, instead of scrolling back and trusting memory. As always, treat what the assistant produces as a draft you read, not output you ship unseen.

General information, not legal or financial advice — check any tool's current pricing, terms, and data handling on your own account before you rely on it.

The part these updates leave alone

Every update this week sharpens the doing — the model that drafts, the layout that renders, the session you can retrace. What none of them touch is the agreement around the work: what was promised, what counts as extra, and when the money actually arrives. That is the seam where projects quietly lose money, and it does not get fixed by a better model or tidier spacing. It gets fixed by writing the deal down and holding people to it — which is the one piece of your stack we would argue you most want to own outright rather than rent loosely.

If our two filters are the checklist you want for the rest of your tools, our note on how to choose freelance management software goes deeper on the same instinct, and privacy-first freelance software is the matching read for the "can I get my stuff out" thread. When a client asks for something that was never in the deal, turning that into a documented, signed decision is exactly what writing a change order covers — and freelance software that locks scope after signing is what we built for the rest.

We run Tool Radar regularly. If there is an adjacent tool you think every freelancer should know about, tell us at ideas@driftcatch.app — we read every message, and good suggestions tend to show up in the next edition.